Using your system
Use your local AI from a browser or a terminal
Your staff can use a shared local AI server through browser chat, while developers connect from a terminal. Open WebUI and OpenCode are two interfaces we can configure to use your local model endpoint.
The model runs on the server; the interface runs where you work
People do not have to sit beside the AI machine to use it. Their browser or terminal sends a request over the network, the server runs the selected model, and the answer comes back to that interface. A shared endpoint lets suitable laptops and desktops use the same model infrastructure.
The interface and the model serve different purposes. The interface handles the conversation and any permitted tools. The model generates the response. Choosing a familiar interface does not change the capabilities of the model behind it, so the combination still needs to be configured and tested.
Browser chat with Open WebUI
Open WebUI provides a browser-based way to chat with a model. It can connect to local runtimes through supported provider protocols. An employee opens the configured address, signs in where authentication is enabled, and selects an available model.
We can configure a browser experience around everyday questions, writing and approved document access. Document search needs its own parsing, indexing and access rules; attaching a chat interface does not automatically connect every company file. Decide which sources should be available and how they stay up to date.
Terminal access with OpenCode
OpenCode offers an interactive coding interface in the terminal. Its provider configuration supports local runtimes and custom compatible endpoints. That lets a developer work with a model hosted on your own server while staying in the tools they already use.
A coding agent may read project files, propose edits or execute commands through its tools. Configure its permissions for the work you intend it to perform, including approval requirements for changes and commands. A private model endpoint does not make every generated command appropriate to run. Review the changes and run the project's normal checks.
Decide which connections are allowed
Both a chat client and a terminal client can be configured with external services. Keep local and cloud models clearly identified, and review additional tools that can send information elsewhere. A cloud model selected in a locally installed application still receives the request in the cloud.
For colleagues away from the office, arrange authenticated access through an agreed network connection. Avoid treating the raw inference port as a public service. Model access, network access and tool permissions are separate controls, each with a role in the installation.
Make the handover useful for the people using it
An installation should leave your team with working client connections, an agreed model selection, access instructions and a support contact. Confirm who administers accounts, where saved conversations live and how model updates are introduced.
You are not restricted to these two interfaces. Other applications can connect when their API and model requirements are supported. The practical advantage is a shared service your team can reach from its existing computers, with tools chosen around the work rather than around the location of the hardware.
Sources & further reading
Official documentation behind the technical details in this guide.